Expired75% confidenceFactTime unknown
Running a million-token context used to require 10 H20-level GPUs, whereas DeepSeek V4 now only needs 1–2.
1
Sources
75%
Confidence
Medium-term (~90 days)
Relevance
6/2/2026
First Seen
Valid until: 8/31/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedAn 8-card H20 configuration has a total video memory of 1128GB, enabling complete loading of ultra-large-scale models such as DeepSeek-V4-Flash without video memory quantization.74% similarPartially VerifiedNVIDIA A100专业训练卡拥有40GB或80GB HBM2e显存，单卡售价超过1万美元67% similarUnverifiedDeepSeek-R1 32B模型约需20GB以上显存，通常需要专业级显卡或多卡并行67% similarUnverifiedUsing the V100 (16GB VRAM) as a reference, the Ornith 9B model at Q6_K requires approximately 7–8GB of VRAM, while the 35B MoE requires approximately 20–22GB and needs dual V100s.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/36600API
curl https://kongchang.com/api/v1/knowledge/claims/36600MCP
get_claim(id=36600)